US 11,989,233 B2
Transformation for fast inner product search on graph
Shulong Tan, Santa Clara, CA (US); Zhixin Zhou, Los Angeles, CA (US); Zhaozhuo Xu, Houston, TX (US); and Ping Li, Bellevue, WA (US)
Assigned to Baidu USA LLC, Sunnyvale, CA (US)
Filed by Baidu USA, LLC, Sunnyvale, CA (US)
Filed on Sep. 27, 2020, as Appl. No. 17/033,791.
Claims priority of provisional application 62/929,751, filed on Nov. 1, 2019.
Prior Publication US 2021/0133246 A1, May 6, 2021
Int. Cl. G06F 16/901 (2019.01); G06F 16/903 (2019.01); G06F 17/16 (2006.01)
CPC G06F 16/9024 (2019.01) [G06F 16/90335 (2019.01); G06F 17/16 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving a query; and
searching a final graph using inner product to identify a set of one or more data elements relevant to the query, in which the final graph was formed by performing steps comprising:
forming a transformed dataset by performing a transformation on a dataset comprising original data elements and by adding an origin element to the transformed dataset;
constructing a proximity graph of the transformed dataset using custom character2 distance, in which at least some of the data elements of the transformed dataset are vertices in the proximity graph and connections between vertices in the proximity graph are edges;
identifying a set of neighbors of the origin element on the proximity graph;
removing the origin element and its edges from the proximity graph; and
obtaining the final graph by replacing the vertices in the proximity graph, which comprises data elements from the transformed dataset, with corresponding original data elements.